Examples on the Application of the

LQI Criterion

JCSS
Joint Committee on Structural Safety

Date: June 2009

Name: M. Schubert, M.H. Faber
Version: Draft version for discussion
JCSS Examples on the Application of the LQI Criterion Page 2

1 Introduction

In the following three examples are provided on the use of the LQI criterion for the
assessment of the conformity of decision alternatives with societal preferences for
investments into life saving activities. In a more usual language this is also understood as an
assessment of whether decision alternatives are societal acceptable. The examples consider
the following decision situations:

Continuous decision alternatives

Discrete decision alternatives

Combinations of discrete decision alternatives

and thereby aim to cover the spectrum of the most usually occurring decision situations in
engineering design and assessment.

2 Design of a Structural Member

Here we consider a typical engineering decision situation where the cross-sectional

dimension of a steel bar under tension loading is to be optimized and assessed for
acceptability. The cross-sectional dimension of relevance, i.e. the cross-sectional area A is
continuous decision parameter. The failure event of the bar is modeled by the event that the
yield stress of the bar is exceeded by the load effect due to the external loading, here assumed
associated with an extreme event of a variable load S . The annual extreme load effect
modeled by a Gumbel distribution with an expected value S 9.50 kN and a standard
deviation S 1.50 kN . The resistance R of the steel bar is given through the cross-sectional
area A and the yield stress of the steel f y . The yield stress of the steel is modeled by a Log-
normal distribution with expected value R 260 MPa and standard deviation R 18.2 MPa .

The annual probability of failure conditional on the cross-sectional area of the steel bar can
thus be calculated as:

Pf A FR ( x | A) f s x dx (1)
0

The LQI criterion can be written as:

dPf ( A) q dC y ( A, t , )
0 (2)
dA C ,d 2%,a N PE k g dA
JCSS Examples on the Application of the LQI Criterion Page 3
In order to proceed it is now necessary to fix the various parameters of Equation (2). It is thus
assumed that the conditional probability of mortality k given the event of failure of the steel
bar is equal to 1. Furthermore it is assumed that 12 persons are exposed to the failure event.

The costs per mm 2 steel cross sectional area are assumed to be equal to 5000 CHF. Assuming
that in the case of failure the steel bar will be systematically renewed to its original condition
the net present value of the construction costs (including future renewal costs) may be
represented by:

C y ( A, t , ) C y*,1 ( A) C *y ,2 ( A) 1 1 ln 1 Pf A t 1
1
(3)

where C *y ,1 ( A), C *y ,2 ( A) correspond to the construction costs and the renewal costs of the steel
bar respectively; here assumed identical equal to C y ( A) . corresponds to a constant annual
discounting rate.

Table 1: Parameters required for the decision optimization.

g Gross Domestic Product per 35931 CHF

capita

C ,d 2%, a Demographic constant 18.9

q - 0.175

k Conditional probability of 1
mortality given failure

N PE Number of exposed persons 12

C *
y A Construction costs
5000 A CHF
b Benefit of structure 1.2 104 CHF / yr
CO Compensation costs 1.8 Mio. CHF
CU Cleaning costs 3 104 CHF
Interest rate 2%
t Time period 100 yr

In Table 1 a summary is provided of the parameters required for the structural optimization
and assessment of acceptance.

The optimal structural design is identified through the following object function:

1 1 1 1
t t

Z A, t , b C y A Pf A C y ( A) k N PE CO CU (4)
ln 1 ln 1
JCSS Examples on the Application of the LQI Criterion Page 4
where it is assumed that the benefit b A associated with the structure is constant equal to
1.2 104 CHF / yr . The failure costs include besides the renewal costs C y ( A) the costs of
cleaning CU as well as compensations for lost lives CO . The annual constant discounting rate
is set equal to 2% and a time period t of 100 years is chosen. Since a long time period is
considered in this example, the discount rate is chosen according to the long term economic
growth rate. Rackwitz (2005) provides a discussion on discount rates.

The optimization of the object function in Equation (4) and the assessment of the LQI
criterion from Equation (2) is illustrated in Figure 1 A) - 1 D).

Based on the LQI criterion it is found that cross-sectional areas smaller than 93 mm2 cannot
be accepted. This corresponds to an annual probability of failure equal to
Pf 5.15 106 [ yr 1 ] , and construction costs equal to 4.65 103 CHF yr 1 .

Figure 1: Summary of the results.

JCSS Examples on the Application of the LQI Criterion Page 5
It is seen that the object function does not attain its maximum in the acceptable region of
decision alternatives. The societal preference for investments into life saving in this case
requires that more is invested into life saving.

3 Protection Against Natural Hazards

In this example the identification of decision alternatives for the protection against natural
hazards is considered. It is assumed that a specific location is considered where people are
exposed to natural hazards through some activity which yields a benefit for society; e.g. a
roadway segment exposed to rock-fall or avalanche hazards. Furthermore, it is assumed that
the decision to establish the roadway segment in the first place in the way it was done was
optimal and in coherency with societal preferences in regard to investments into life saving
activities at the time.

The problem complex investigated in the further is related to whether additional protective
measures which have become available since the roadway segment was established should be
pursued from the perspective of societal acceptance as well as from the perspective of
economical optimization.

At the location of interest the natural hazard of relevance occurs with an annual rate equal to
p0 0.035 . The conditional probability of mortality k p0 given the event of the natural
hazard is assumed equal to 0.40 and the expected value of the number of exposed persons is
N PE p0 2.0 .

The investigated five different decision alternatives are denoted by pi , i 1, 2,..5 . Depending
on the chosen decision alternative the event of the natural hazard is reduced to pi and the
number of exposed persons is changed to N PE ,Pi .

An analysis of the effect and costs of the different decision alternatives has resulted in the
results which are summarized in Table 2.

Table 2: Parameters for the different decision alternatives.

p i
C *y ( pi ) C y ( pi , t , ) CU , pi N PE , pi k pi *p
Option
i

yr 1 CHF CHF yr 1 CHF CHF CHF yr 1

p0 3.5 102 - - 0.5 103 2 0.4 -

p1 2.0 102 5.0 10 3
9.4 10 2
1.0 103 4 0.8 8.7 10 3
p2 2.0 103 1.5 105 1.6 103 1.5 103 5 0.3 1.8 10 2
p3 5.0 103 3.0 106 3.7 104 1.9 103 10 0.4 4.6 103
p4 5.0 104 2.0 105 2.0 103 2.0 103 8 0.3 1.110 2
p5 1.0 104 1.0 106 1.0 104 7.0 103 10 0.2 1.3 10 2

Based on the LQI criterion it is now assessed which of the different decision alternatives are
acceptable. The LQI criterion yields:
JCSS Examples on the Application of the LQI Criterion Page 6

C y ( pi , t , ) q
N PE p0 k p0 ( p0 )
C ,d 2%,a g
0 ( pi ) * ( pi ) (5)
N PE pi k pi

With the annual costs of a given decision alternative given as:

C y pi , t , C *y pi C *y pi 1 1
t
ln 1 1

pi t 1 (6)

And parameters as given in Table 1 and Table 2.

From Equation (5) it is found that as long as * pi larger is than pi the decision
alternative i is conform with societal preferences for life saving and must thus be accepted.
For the present example this is the case for decision alternatives p2 , p4 and p5 .

As it is seen that three of the decision alternatives are acceptable the question is which of
these are the optimal, considering a maximization of life-cycle benefits. The object function
to be optimized can be written as:

Z pi , t , B pi , t , C *y pi R pi , t , (7)

in which the benefit of the individual decision alternatives B pi , t , are given through the
reduction in life risks and can be determines as:

1 1
t

B pi , t , 0 k0 N PE ,0 pi k pi N PE , pi CO
ln 1
(8)

The remaining risk is calculated by:

1 1
t

R pi , t , pi C y pi CU pi (9)
ln 1

In Figure 2 a LQI-diagram is shown (Nathwani et al. (1997)). The values on the y-axis are
proportional to the change in the expected number of fatalities N PE , pi k pi v pi with
v pi v pi v p0 . The proportionality factor for the y-axis is g . The values on the x-axis are
proportional to the annual costs of a given decision alternative C y pi . The proportionality
factor for the x-axis is the q Cx1 . The shaded area in Figure 2 illustrates the region where
decisions are not acceptable, i.e. N PE , pi k pi v pi C y pi q Cx1 . The results for the
investigated decision alternatives p1 p5 are plotted in Figure 2.

From the LQI-diagram it is clearly seen that the decision alternatives p1 and p3 are not
acceptable. Only the options p5 , p4 , p2 can be accepted. Table 3 summarizes the results of the
assessment of decision alternatives.
JCSS Examples on the Application of the LQI Criterion Page 7

Table 3: Overview of assessment result for the investigated decision alternatives.

Option pi * pi B pi , t , C *y pi R pi , t , Z pi , t ,
p1 2.0 102 8.7 10 3 2.82 105 5.0 103 4.44 104 3.76 105
p2 2.0 103 1.8 10 2 1.96 105 1.5 105 1.32 104 3.27 104
p3 5.0 103 4.1 10 3 6.27 104 1.0 106 6.53 105 3.59 106
p4 5.0 104 1.1102 2.1105 2.0 105 4.4 103 5.58 102

p5 1.0 104 1.3 10 2 2.18 105 1.0 106 4.83 103 7.87 105

It is interesting to note that decision alternative p2 maximizes the life-cycle benefit even
though the decision alternative p5 yields a lower life risk.

Figure 2: LQI-Diagram for the five considered decision alternatives.

4 Protection Against Natural Hazards

As a final example consider the assessment of decision alternatives comprised through

packages of discrete decision alternatives. This would correspond to the foregoing example
but where now new decision alternatives are constructed by combination of the previous. It is
assumed that the initial life risk is given as p N PE , p k p p 3.5 103 yr 1 . Five different
0 0 0 0

decision alternatives to reduce the risks are to be combined and assessed in regard to
optimality and acceptability. The annual costs C y , p and the risk reductions p
i i

corresponding to the different alternatives are summarized in Table 4.

It is further assumed that the different decision alternatives can be freely combined and that
their combined effect on risk reduction as well as their costs can be assessed by simple
addition. As a first step the alternatives are ranked in accordance with their efficiency. The
JCSS Examples on the Application of the LQI Criterion Page 8
most efficient decision alternative is p1 , as this represents the largest ration between p and i

C y , pi . The next most efficient decision alternative is p2 . The ranked list of decision
alternatives is provided in column 5 of Table 4.

Table 4: Input parameters for the assessment of packages of measures.

Alternative pi C y , pi pi C y , pi Rank
1
[ yr ] [CHF yr ] 1
[CHF ]1
[ ]
2 6
p1 2.0 10 3 10 3
6.7 10 1
p2 2.0 103 2 103 1.0 106 3
p3 8.0 103 7 103 1.1106 2
p4 4.0 103 9 103 4.4 107 4
p5 8.0 104 1 10 4 8.0 108 5

The combinations of packages of risk reducing measures to be investigated can now simply
be established through the simple successive addition of alternatives in accordance with the
rank number. In this way first combination p1 p2 is assessed, thereafter p1 p2 p3 , etc.
Additional risk reducing measures, following the ranked list, should be pursued as long as the
acceptance criterion for packages is fulfilled:

g
C y ( pi | i ) Cx N PE pi | i k pi | i ( pi | i ) 0
q
g
C y ( pi | i ) Cx ( pi | i ) 0 (10)
q
i p1 , p2 ,..., pi

Table 5: Calculation of the acceptability of the combination.

g
Combination i Alternative pi C y pi | i C y ( pi | i ) Cx ( pi | i )
q
[ ] [ yr 1 ] [ yr 1 ]
p1 p1 2.0 102 2.0 102 7.5 10 4
p1 , p3 p3 8.0 103 8.0 103 2.4 103
p1 , p3 , p2 p2 2.0 103 2.0 103 5.8 103
p1 , p3 , p2 , p4 p4 4.0 103 4.0 103 6.5 103
p1 , p3 , p2 , p4 , p5 p5 8.0 104 8.0 104 6.9 10 4

The result is given in Table 5. For the present example this implies that all decision
alternatives p1 p4 should be implemented. For p5 the acceptance criterion is no longer
fulfilled. This may be visualized by the indifference curve for the LQI shown in Figure 3.
JCSS Examples on the Application of the LQI Criterion Page 9
The dashed lines in Figure 3 illustrates the LQI indifference curves. Any decision which
simply corresponds to expenditures (or gains) and life savings (losses) for which the
corresponding LQI remains the same are LQI indifferent; corresponds to a translation in the
Figure along the LQI indifference curves. The direction indicated on the LQI indifference
curves show in which direction the LQI increments positively. The actual value of the LQI is
simply read from the given location in the plot. The most efficient risk reducing measure with
the highest efficiency yields the largest increment in LQI, i.e. alternative p1 , the second most
efficient alternative is p3 . The incremental benefit of the implementation of additional risk
reducing measures is seen to approach the gradient of the LQI indifference curve. For the last
decision alternative it is seen that the increment is seen to be smaller the gradient of the LQI
indifference curve and this alternative should thus not be implemented. If this alternative (i.e.
p5 ) would be implemented this would lead to a societal loss; the benefit is smaller than the
expenditure. The maximal societal benefit achievable from the implementation of the
packages of decision alternatives is illustrated through d max , see Figure 3.

Figure 3: LQI indifference curves and LQI increase associated with the combined decision
alternatives p1 , p1 ,.., p5 .

The foregoing only addressed the assessment of the conformity of decision alternatives with
societal preferences in regard to life saving investments. The assessment of the economical
optimality of the investigated packages should of course also be undertaken. This is however,
not addressed in this example but follows straightforwardly following the same principle as
for individual discrete decision alternatives, see also the preceding example.

References
Nathwani, J. S., Lind, N. C. & Pandey, M. D. (1997) Affordable Safety by Choice: The Life
Quality Method, Waterloo, Ontario, Canada, Institute for Risk Research, University of
Waterloo.

Rackwitz, R. (2005) The Philosophy Behind the Life Quality Index and Empirical
Verifications. Memorandum to JCSS. Munich, Germany.